PoliEcho/asm-game-of-life
PoliEcho/asm-game-of-life
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 2 Wiki Activity

fix memory_set function

Browse Source
This commit is contained in:
PoliEcho 2025-07-09 19:23:13 +02:00
parent c0bd3fcbc3
commit 2b8236ee91
1 changed files with 38 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

51
src/str.asm
View File

@ -80,21 +80,33 @@ global memory_set:
memory_set:; takes destination in rdi, byte in sil and lenght in rdx memory_set:; takes destination in rdi, byte in sil and lenght in rdx
; first check if value is 16 byte alligned ; first check if value is 16 byte alligned
mov r9, rdi xor r8, r8
mov rax, rdi mov r9, rdi; move destination to r9
and rax, 0xF; offset is stored in rax
mov r11, 0x0101010101010101; to extend across whoule register mov r11, 0x0101010101010101; to extend across whoule register
movzx rsi, sil movzx rsi, sil
imul r11, rsi; to extend across whoule register imul r11, rsi; to extend across whoule register
cmp rdx, 16
jnl .write_16_or_more_bytes
mov r8b, dl
jmp .write_less_than_16_bytes
.write_16_or_more_bytes:
mov rax, rdi; move destination to rax
and rax, 0xF; offset is stored in rax
test al, al; check if resault is 0 test al, al; check if resault is 0
jz .addr_is_16_Byte_alligned jz .addr_is_16_Byte_alligned
mov rax, r11
mov r8b, 16 mov r8b, 16
sub r8b, al; now offset to first higher 16 byte alligned address is stored in r8 sub r8b, al; now offset to first higher 16 byte alligned address is stored in r8
mov rax, r11
.write_less_than_16_bytes:
sub rdx, r8; we will write these bytes now sub rdx, r8; we will write these bytes now
;add rdi, rdx ;add rdi, rdx
@ -102,30 +114,35 @@ memory_set:; takes destination in rdi, byte in sil and lenght in rdx
mov rcx, 1; we will allwais copy only once mov rcx, 1; we will allwais copy only once
.check_qword:; check if offset is more than qword
cmp r8b, 8 cmp r8b, 8
jl .check_dword jl .check_dword
rep stosq rep stosq
sub r8b, 8
.check_dword: .check_dword:
cmp r8b, 4 cmp r8b, 4
jl .check_word jl .check_word
rep stosd rep stosd
sub r8b, 4
.check_word: .check_word:
cmp r8b, 2 cmp r8b, 2
jl .check_byte jl .check_byte
rep stosw rep stosw
sub r8b, 2
.check_byte: .check_byte:
test r8b, r8b; check if offset is 1 or 0 test r8b, r8b; check if offset is 1 or 0
jz .addr_is_16_Byte_alligned jz .addr_is_16_Byte_alligned
rep stosb rep stosb
dec r8b
.addr_is_16_Byte_alligned: .addr_is_16_Byte_alligned:
shr rdx, 4; set it to how many 128bit(16Byte) chunk we need mov rcx, rdx
shr rcx, 4; set it to how many 128bit(16Byte) chunk we need
test rcx, rcx; check if we need to write aditional 16 bytes at all
jz .function_exit
%ifdef AVX512 %ifdef AVX512
vpbroadcastq xmm8, r11 vpbroadcastq xmm8, r11
%else %else
@ -136,10 +153,18 @@ memory_set:; takes destination in rdi, byte in sil and lenght in rdx
.move_16_bytes: .move_16_bytes:
movdqa [rdi], xmm8 movdqa [rdi], xmm8
add rdi, 16 add rdi, 16
dec rdx sub rdx, 16
test rdx,rdx; test if rdx is zero cmp rdx, 16; test if rdx is less than 16
jnz .move_16_bytes jge .move_16_bytes
.function_exit:
test rdx, rdx; test if rdx is 0
jz .true_function_exit
mov r8b, dl
jmp .write_less_than_16_bytes
.true_function_exit:
mov rax, r9; return pointer to memory area same as memset in libc mov rax, r9; return pointer to memory area same as memset in libc
ret ret